Analysis Runner Helper

GitHub Link to Code.

Analysis runner helper for feature importance operations.

This module provides helper methods for running feature importance analyses on individual sub-comparisons, extracting common logic from FeatureImportanceManager methods.

class mdxplain.feature_importance.helper.analysis_runner_helper.AnalysisRunnerHelper

Helper class for running feature importance analyses.

Provides static methods for executing feature importance analysis on individual sub-comparisons and processing the results. These methods extract common logic from FeatureImportanceManager to improve code organization and reusability.

Examples

>>> # Run analysis on single sub-comparison
>>> result = AnalysisRunnerHelper.run_single_analysis(
...     analyzer_type, X, y, sub_comp
... )

>>> # Process and store analysis result
>>> AnalysisRunnerHelper.store_analysis_result(
...     fi_data, result, metadata
... )

static store_analysis_result(fi_data: FeatureImportanceData, result: Dict[str, Any], metadata: Dict[str, Any]) → None

Store analysis result in FeatureImportanceData object.

Takes the analysis result and metadata and stores them properly in the FeatureImportanceData container. Enriches metadata with model and analysis metadata for downstream visualization.

Parameters

fi_dataFeatureImportanceData: Feature importance data container to store result in
resultDict[str, Any]: Analysis result from analyzer containing importances, model, metadata
metadataDict[str, Any]: Metadata dictionary describing the analysis

Returns

None: Stores result in the fi_data object

Examples

>>> AnalysisRunnerHelper.store_analysis_result(
...     fi_data, analysis_result, metadata_dict
... )

static run_comparison_analysis(pipeline_data: PipelineData, comp_data: ComparisonData, analyzer_type: AnalyzerTypeBase, analysis_name: str) → FeatureImportanceData

Run analysis on all sub-comparisons in a comparison.

Processes all sub-comparisons within a comparison object, running the specified analyzer on each one and collecting results.

Parameters

pipeline_dataPipelineData: Pipeline data object containing data and comparisons
comp_dataComparisonData: Comparison data object containing sub-comparisons
analyzer_typeAnalyzerTypeBase: Analyzer instance to use for all sub-comparisons
analysis_namestr: Name for the analysis (for metadata)

Returns

FeatureImportanceData: Complete feature importance data with all sub-comparison results

Examples

>>> fi_data = AnalysisRunnerHelper.run_comparison_analysis(
...     pipeline_data, comp_data, analyzer, "my_analysis"
... )
>>> print(len(fi_data.data))  # Number of sub-comparisons analyzed